I used to have this 65c816 fantasy. That Atari or Commodore had put out a machine in this lineage rather than jumping to the 68000. On paper it looks appealing -- 6502 compatibility, 24-bit address bus, super fast interrupt responsiveness, very cycle efficient.
So I wasted some time last year futzing around on the C256 Phoenix project -- wrote an emulator for it (or some aspect of what it looked like on paper) since I didn't have a board yet, wrote some code for it, had a falling out with the creator, a bit of a saga. In any case, what I learned in this process is that I really don't like the '816. It was a hack bolted onto the side of the 6502. Bill Mensch couldn't or wouldn't add new addressing modes and register sets to the instruction set to allow 8-bit and 16-bit to co-exist, so there's a constant awkward mode switch back and forth in terms of how the registers are viewed. Even worse, it has a 24-bit address bus but no registers can deal with 24-bit pointer values, so a constant awkward play with using indirect addressing or using multiple instructions to modify addresses or switch pages and so on. And then on top of that the physical chip doesn't break out all 24 address lines, so interfacing the thing is a pain.
The 816 is still produced 35 years later, and available in PLCC and SMT form factors. But they've had this 35 years to make a version of the chip that breaks out all the address lines instead of multiplexing it, which would make sense with the higher pin count available in SMT. But they haven't done it. Why? They went through the effort of making a microcontroller version of it that actually does break out all 24 lines, which nobody really uses. But no MPU.
In the end the 816 was a bit of a hack done especially for Apple. I understand why Mensch did it, but they could have done better.
In contrast, the Z80 has continued to evolve. You can get a brand new eZ80 etc. with large address space and higher clock speeds and excellent cycle efficiency, etc. and not have to muck around with demultiplexing the address bus, etc.
So all said I now understand why engineers at Atari and Commodore and Apple said "f it" about the 6502 and switched to the 68000. The real tragedy of the late 80s/90s era was that Motorola pissed away the excellency of 680x0 by jumping first to the 88000 (failure) and then the PowerPC (less of a failure but not a success for Motorola). Meanwhile Intel kept backwards compatibility and squeezed every ounce it could out of the existing x86, and that's one reason they did better and Wintel maintained dominance.